Fundamental Concept for Decision Support

The closed loop on the left describes an interaction model between the SAF applications and users for real-time problem analysis. There are at least four decision-supporting steps, including:

The closed loop on the right describes an interaction model between SAF applications and domain experts to improve knowledge content and provide subsequent feedback to the closed loop on the left. This provides consistent and constantly improved decision support for a user. The closed loop on the right is a design-time process that is typically operated as a background task. The steps to improve knowledge content include:

This basic closed-loop concept of decision support can be applied to various SAF components (for example, a learning engine for refinement) to provide real-time interactions and design-time optimization for each component.

The architecture designed to achieve the basic concept is illustrated below. The SAF architecture follows a layered design and a componentized design. A layered design provides vertical decoupling of user interface, service API, and engines, while a componentized design provides an open integration framework for unrestricted extension, for example, multiple knowledge bases, multiple search engines. The architecture is designed specifically to make sure components can be reused for different business use cases, so a business application can be configured by simply reassembling a set of building blocks.

In the figure below, SAP CRM is used as an example to illustrate the architecture when associated with data sources (such as SDB) and business applications (such as IC WebClient knowledge search).

Architecture Overview

The SAF is composed of a knowledge repository, search engine adaptors, and a learning engine. Each component has a set of programming APIs that are created for both runtime and design-time services. The runtime services are used by business applications, and the design-time services are mainly for performance improvement.